A Hybrid Deep Learning Approach With Multimodal Neurocognitive Features and Quantum Optimization for Early Detection of Alzheimer's Disease and Related Dementias

ABSTRACT

Dementia is a progressive condition, often associated with Alzheimer's disease that gradually impairs cognitive functions and affects the ability to carry out simple daily tasks. A major challenge in diagnosing these neurological disorders is that their symptoms often overlap with normal aging or other brain‐related diseases, especially in the early stages, which complicates clinical diagnosis and effective treatment. Existing assessment methods, such as brain imaging, memory assessments, and intelligent questioning systems (IQS), are frequently used but often insufficient for monitoring progression. With the emergence of artificial intelligence (AI) and other learning frameworks, clinical trials have been enhanced by the availability of definitive outcomes, offering a promising avenue for predicting treatment outcomes. In this research, a novel ensemble deep learning (DL) framework is proposed that integrates multimodal datasets to detect dementia in individuals effectively. The proposed architecture introduces innovative convolutional‐transformer networks for MRI image analysis and global gated attention networks for analyzing different text vectors. Additionally, improved quantum marine predator (IQMP) optimization is used to tune the hyperparameters of the classification networks. The complete evaluation of the proposed model was conducted using multiple heterogeneous datasets, including the OASIS MRI image dataset and the Dementia Benchmark datasets, to assess the networks' efficacy. To demonstrate the performance of the proposed approach, metrics such as accuracy, precision, recall, specificity, and F 1 score were evaluated and compared with those of other state‐of‐the‐art learning frameworks. The findings indicate that the proposed model achieved an accuracy of 0.99, a precision of 0.989, a recall of 0.988, a specificity of 0.99, and an F 1‐score of 0.98, demonstrating its effectiveness in diagnosing the early stages of dementia. The experimental results suggest substantial potential over current state‐of‐the‐art techniques. In this research, a novel ensemble deep learning framework is proposed that integrates multimodal datasets to enable effective early detection of dementia.